The Vulnerability of Kentucky s Manufacturing Economy to Increasing Electricity Prices. Aron Patrick

Transcription

1 The Vulnerability of Kentucky s Manufacturing Economy to Increasing Electricity Prices Aron Patrick Kentucky Energy and Environment Cabinet Department for Energy Development and Independence October, 2012 energy.ky.gov

2 This page has been intentionally left blank.

3 Executive Summary Kentucky s low electricity prices have fostered the single-most electricity-intensive manufacturing economy in the United States, a manufacturing economy that is now threatened by future electricity price increases. This study builds upon the notion that low energy costs are a catalyst for commercial growth by quantifying the specific vulnerability of the largest economic sectors of the Commonwealth, in terms of total employment, to future electricity price increases. Using a statistical analysis technique called multiple regression of panel data with fixed effects, this study modeled the responsiveness of employment across the United States to changes in the price of electricity from 1990 to 2010 for the top five employment sectors in Kentucky: manufacturing, retail services, hospitality, healthcare, and government. Elasticities were developed for each of these economic sectors to calculate changes in employment, given a specific change in the price of electricity, and can be generally applied to the 48 contiguous United States. Given a 25% forecasted increase in the real price of electricity in Kentucky between 2011 and 2025, this study estimates the Commonwealth will likely lose, or fail to create, approximately 30,000 fulltime jobs in the long-term. Manufacturing establishments were found to be most responsive to changes in electricity prices and can be expected to permanently shed 17,500 full-time jobs. The other largest employment sectors in Kentucky, retail stores, restaurants, and hotels, were less than half as responsive as the manufacturing sector to increasing electricity prices, and combined, can be expected to fail to create 12,500 full-time jobs. However, in the fourth and fifth largest employment sectors, healthcare and government, no statistically significant relationship could be identified between electricity prices and total employment. While total employment in Kentucky is expected to continue to rise in other sectors, the Commonwealth should develop strategies to mitigate vulnerability to energy price increases, volatility, and risk exposure. Additionally, Kentucky should maintain focus on education and workforce development in emerging industries that are less reliant on energy-intensive manufacturing processes. These forecasted electricity price increases, in addition to the current trend towards off-shoring and automation of manufacturing processes, have the potential to transform the economies of manufacturing states like Kentucky. energy.ky.gov i

4 This page has been intentionally left blank. energy.ky.gov ii

5 The Vulnerability of Kentucky s Manufacturing Economy to Increasing Electricity Prices Table of Contents Executive Summary Table of Contents List of Figures List of Tables i iii iv iv Background: Kentucky s Electricity-Intensive Manufacturing Economy 1 Business Response Options 7 Findings 7 Manufacturing Employment 9 Retail Trade Employment 10 Hospitality Employment 11 Healthcare Employment 12 Government Employment 12 Conclusions 12 Statistical Appendix 13 Data Analyzed 13 Analytical Method 14 Complete Models 16 Model Diagnostics 18 Acknowledgements 20 References 21 energy.ky.gov iii

6 List of Figures Figure 1: Kentucky Electricity Consumption by Economic Sector, Figure 2: United States Electricity Consumption by Economic Sector, Figure 3: Kentucky Electricity Consumption by Economic Sector, Figure 4: United States Electricity Consumption by Economic Sector, Figure 5: Total Real Electricity Prices, Kentucky vs. the United States, Figure 6: Electricity-Intensity of Production, Kentucky vs. the United States, Figure 7: Kentucky Gross Domestic Product by Economic Sector, Figure 8: Kentucky Employment by Economic Sector, Figure 9: Kentucky Electricity Intensive Employment Forecast, Figure 10: Kentucky Manufacturing Employment Forecast, Figure 11: Kentucky Retail Trade Employment Forecast, Figure 12: Kentucky Hospitality Employment Forecast, Figure 13: Model of Manufacturing Employment Predicted vs. Observed Values 18 Figure 14: Model of Manufacturing Employment Normal Q-Q Diagnostic Plot 18 Figure 15: Model of Retail Trade Employment Predicted vs. Observed Values 18 Figure 16: Model of Retail Trade Employment Normal Q-Q Diagnostic Plot 18 Figure 17: Model of Food & Accommodation Employment Predicted vs. Observed Values 19 Figure 18: Model of Food & Accommodation Employment Normal Q-Q Diagnostic Plot 19 Figure 19: Model of Healthcare Employment Predicted vs. Observed Values 19 Figure 20: Model of Healthcare Employment Normal Q-Q Diagnostic Plot 19 Figure 21: Model of Government Employment Predicted vs. Observed Values 19 Figure 22: Model of Government Employment Normal Q-Q Diagnostic Plot 19 List of Tables Table 1: National Manufacturing Sector Electricity-Intensity and Kentucky Employment 4 Table 2: Multiple Regression Models of Electricity Prices & Employment by Sector 16 Table 3: Multiple Regression Models of Electricity Prices with Robust Standard Errors 17 energy.ky.gov iv

7 Kentucky s Energy-Intensive Economy In 2011, 49% of all electricity consumed in Kentucky went to industrial users, compared with 26% for the United States as a whole, as illustrated in Figures 1 and 2 below. The reason for this is obvious industries requiring large amounts of electricity for production have an incentive to locate in states where they can anticipate that electricity costs will remain low. The industrial nature of Kentucky s electricity load is by no means a recent development. Ever since the first power plants were built in the Commonwealth, most of the electricity produced went to large factories. Over the past 50 years for which there is reliable data, industrial users have consumed an average of 60% of all electricity generated in Kentucky annually, as illustrated in Figure 3 below. These proportions for the United States as a whole have historically been far more balanced, as illustrated in Figure 4 below. Figures 1 & 2: Electricity Consumption by Economic Sector, Kentucky vs. the United States, 2011 Figures 3 & 4: Electricity Consumption by Economic Sector, Kentucky vs. the United States, energy.ky.gov 1

8 Coal has historically provided the Commonwealth both low-cost electricity and energy security. Nominal electricity prices in Kentucky have increased since 1970 at about 2% annually, which is less than the average rate of inflation during this same period. When adjusted for inflation, 1 as illustrated in Figure 5 on page 3, real electricity prices actually fell in Kentucky from 1980 to 2003, and have risen over the past decade with increases in the price of all fossil fuels. Since 1992, Kentucky has maintained one of the lowest four electricity prices in the nation, running neck and neck with the coal and hydroelectric states of Idaho, Wyoming, Washington, and West Virginia. Figure 6 on page 3 illustrates that Kentucky is home to the most electricity-intensive economy in the United States. Simply stated, this means that Kentucky industries use more kilowatt-hours of electricity to produce one dollar of GDP than any other state and are, therefore, more sensitive to changes in electricity prices than any other state. In 2009, the most-electricity-intensive sectors nationally were aluminum smelting, iron & steel mills, paper mills, chemical production, and glass manufacturing, which required on average between 0.5 and 4.5 kilowatt-hours of electricity to produce $1 worth of goods. At current Kentucky industrial electricity prices, each dollar of shipments from these industries required between $0.025 and $0.222 worth of electricity. In other words, up to a quarter of total revenues in these industries go to electricity costs. In Kentucky, the most-intensive of these manufacturing processes, which require more than 0.5 kilowatthours of electricity to produce $1 of goods, directly contributed $5 billion, or 3.2%, to the Commonwealth s total 2009 GDP and employed 12,685 Kentuckians. 2 The national average electricityintensity of each NAICS manufacturing sector present in Kentucky is summarized in Table 1 on page 4 along with the total number of employees and the contribution of each industry to Kentucky s 2009 State GDP based on data provided by the U.S. Census Bureau s Annual Survey of Manufactures and the U.S. Bureau of Economic Analysis. 3 This table provides an approximate rank ordering of sensitivity to electricity prices between types of manufacturing operations present in Kentucky. energy.ky.gov 2

9 Figure 5: Total Real Electricity Prices, , Kentucky vs. the United States Real Electricity Price, (2010 Cents per kwh) Total Real Electricity Prices, Kentucky vs. the United States Other States US Average Kentucky Kentucky Energy Database, EEC-DEDI, 2011 Figure 6: Total Electricity Intensity of Production, , Kentucky vs. the United States Total Electricity Intensity of Production, Kentucky vs. the United States Kilowatt Hours Per Real 2010 Dollar of GDP Other States US Average Kentucky Kentucky Energy Database, EEC-DEDI, 2011 energy.ky.gov 3

10 Table 1: National Manufacturing Sector Electricity-Intensity and Kentucky Employment by NAICS, 2009 NAICS 4 NAICS Description National Electricity Intensity of Production (kwh per $ of Shipment) Kentucky Kentucky Kentucky Average Workers Production Worker Hours (1,000) Value added ($1,000) 3313 Aluminum Production & Processing ,482 6,930 1,083, Iron & Steel Mills & Ferroalloy ,954 6, , Pulp, Paper, & Paperboard Mills ,192 2,382 1,142, Basic Chemical ,043 6,000 2,245, Glass & Glass Product ,015 4, , Foundries ,595 3, , Resin, Syn Rubber, & Artificial Syn Fibers & Filaments ,845 3, , Cement & Concrete Product ,688 2, , Other Nonmetallic Mineral Product ,352 82, Fabric Mills , Coating, Engraving, Heat Treating, & Allied Activities ,434 62, Plastics Product ,552 19,369 1,369, Beverage ,941 3, Sawmills & Wood Preservation ,743 3, , Other Electrical Equipment & Component ,237 2, , Forging & Stamping ,462 2, , Rubber Product ,161 2, , Animal Slaughtering & Processing ,233 17,208 1,126, Fruit & Vegetable Preserving & Specialty Food ,214 6, , Bakeries & Tortilla ,018 6, , Converted Paper Product ,636 10,950 1,167, Semiconductor & Other Electronic Component ,315 44, Spring & Wire Product ,359 4, , Motor Vehicle Parts ,660 31,037 2,942, Other Chemical Product & Preparation , , Printing & Related Support Activities ,092 15, , Machine Shops, Turned Product, & Screw, Nut, & Bolt ,772 5, , Other Fabricated Metal Product ,699 4, , Other Wood Product ,764 10, , Boiler, Tank, & Shipping Container , , Engine, Turbine, & Power Transmission Equipment ,209 2, , Metalworking Machinery ,331 2, , Petroleum & Coal Products , Household & Institutional Furniture & Kitchen Cabinet ,597 2, Dairy Product ,531 3, , Aerospace Product & Parts ,257 2, , Office Furniture (Including Fixtures) ,017 2, Other Miscellaneous ,006 3, , Household Appliance ,576 2, Other General Purpose Machinery ,307 6, , Other Food ,570 2, , Paint, Coating, & Adhesive , , Ship & Boat Building , Ventilation, Heating, Ac, & Commercial Refrigeration ,071 3, , Architectural & Structural Metals ,402 6, , Electrical Equipment ,107 1, , Agriculture, Construction, & Mining Machinery ,407 2, , Medical Equipment & Supplies ,242 2, , Motor Vehicle Body & Trailer ,622 76, Soap, Cleaning Compound, & Toilet Preparation , , Tobacco , Motor Vehicle ,384 22,724 energy.ky.gov 4

11 Figure 7: Kentucky Gross Domestic Product by Economic Sector, Figure 8: Kentucky Employment by Economic Sector, 2009 energy.ky.gov 5

12 Kentucky's electricity-intensive manufacturing economy is threatened by increasing electricity prices. While the price of electricity is only one of several factors influencing industrial location decisions, Kentucky's historically low and stable electricity prices have fostered the most electricity-intensive economy in the United States. In the twenty-first century, the bulwark of the Kentucky economy is clearly manufactured goods the Commonwealth s single largest source of economic activity. Even midrecession, as illustrated in Figures 6 and 7 on page 5, manufacturing in Kentucky accounted for more than $26.6 billion in 2009, or 17% of State GDP, and directly employed 213,330 Kentuckians 2.5 times more than were employed as farmers and 11 times more than were employed as coal miners. In addition to being Kentucky s largest source of revenue and a leading source of employment, manufacturing is sui generis, fulfilling a unique economic function in that most goods are exported, bringing revenue to the Commonwealth from other economies. This is in contrast to the other top employment opportunities in Kentucky: retail services, health care, local government, food service, and construction, which principally depend upon local sources of revenue. Employment opportunities in manufacturing pay more than the two larger employment sectors, retail and hospitality. Large manufacturers, such as General Electric, Toyota, and Ford Motor in Kentucky, also have a more significant multiplier effect on a regional economy because they encourage suppliers to collocate with manufacturing facilities. 5 And this may well be the greatest significance of coal for the Commonwealth: not the number of persons employed in coal mining operations, nor the direct revenue generated from coal exports, but rather the sheer size of the manufacturing industry that has located in Kentucky because of low energy costs. A variety of econometric studies 6,7 have been conducted to estimate the relationship between electricity prices and employment, also finding that increased electricity prices are associated with reductions in employment. However, none of these studies have taken into account the regional disparities in both the forecasted electricity price increases as well as distribution of electricity-intensive manufacturing as a percentage of total employment or state gross domestic product (GDP). Furthermore, none of these existing studies have specifically analyzed the impact of increasing prices on the most relevant employment sectors in the Commonwealth of Kentucky: manufacturing, retail, hospitality, healthcare and government. A 2011 report prepared for the Kentucky state government found that increases in the price of electricity are associated with decreases in overall levels of employment. Specifically, the authors posit that a onetime increase of 25% in the price of electricity would reduce the long-run growth rate in total employment from an average of 3.0% to 2.49% per annum. 8 This current study builds upon the their work by using sector-specific employment as the dependent variable rather than total employment in all sectors to identify particular vulnerabilities within the Kentucky economy. Beyond absolute price, the mere presence of price volatility may make it difficult for electricity-intensive manufacturing businesses to plan ahead and may also discourage capital investment in these engines of economic growth. Electricity price volatility could be included as an independent variable in future studies. For example, one could surmise that during a period of electricity price increases, companies would leave or not expand their existing operations, and this would not necessarily be recovered during periods of declining electricity prices. energy.ky.gov 6

13 Business Response Options to Increasing Electricity Prices Faced with increasing electricity prices, energy-intensive businesses have the following response options. 1. Pass the price increase directly to consumers, in non-competitive markets. 2. Ignore the price increase and accept a reduction in profit margins. 3. Implement energy efficiency measures to lower total electricity consumption. 4. Substitute electricity with alternative energy sources, where available and competitively priced. 5. Seek government incentives or intervention. 6. Implement efficiency in other areas, including labor costs. 7. Relocate to an area where costs of production will be lower. 8. Close. Option 1, passing the price increases directly to product end users, will only be a viable option if that industry has a captive or non-competitive market. If market competition is tight or if there are already lower-cost alternatives available to consumers, manufacturers may have limited room to increase prices. Electricity-intensive industries will not likely be able to choose option 2, since electricity expenditures are such a significant portion of their costs of doing business. In such cases, businesses have probably also already implemented energy efficiency measures, option 3, to increase profit margins. However, as much as possible, more efficient use of electricity is preferable under most conditions. The use of energy substitutes, option 4, for energy-intensive industries in Kentucky may mean substituting direct natural gas combustion for electricity. However, natural gas price volatility, supply, and pipeline access may be prohibiting factors to large scale natural gas substitution. Businesses may also turn to government to either subsidize increasing electricity costs or offset them through taxpayer or ratepayer-funded incentives, option 5. Indeed, many other state governments already offer such incentives to electricity-intensive industries; however, in practice, the long-term affordability of such subsidies must be part of the government s evaluation criterion. Whenever a business chooses options 6, 7, or 8, there should be a negative impact on total employment. Options 7 and 8 could be measured in total number of employees, whereas option 6 would be better measured using total labor hours or wage data. Findings This study builds upon the notion that low energy costs are a catalyst for commercial growth by quantifying the precise vulnerability of the largest economic sectors of the Commonwealth, in terms of total employment, to future electricity price increases. Using a statistical analysis technique called multiple regression of panel data with fixed effects, discussed in greater detail in the Statistical Appendix on pages 13 to 19, this study modeled the responsiveness of employment across the United States to changes in the price of electricity from 1990 to 2010 for the top five employment sectors in Kentucky: manufacturing, retail services, hospitality, healthcare, and government. Elasticities were developed for each of these economic sectors to calculate changes in employment, given a specific change in the price of electricity, and can be generally applied to the 48 contiguous United States. energy.ky.gov 7

14 Figure 9: Kentucky Electricity Intensive Employment Forecast, Given the potential cumulative increase of 25% in real electricity prices between 2011 and 2025, this multiple regression model estimates that Kentucky will likely lose, or fail to create, 30,000 full-time jobs long-term. Manufacturing establishments were the most vulnerable to electricity price increases and can be expected to permanently shed 17,500 full-time jobs. Evidence suggests that, once lost, similar manufacturing employment opportunities will never return. The relative extent of this finding is intuitive given that there are 12,685 jobs in the most-electricity intensive manufacturing sectors alone. Retail stores, restaurants, and hotels were less than half as responsive as the manufacturing sector to increasing electricity prices, and combined, can be expected to fail to create 12,500 full-time jobs. However, in the fourth and fifth largest employment sectors, healthcare and government, no statistically significant relationship between electricity prices and total employment could be identified. The employment forecast illustrated in Figure 9 above is an aggregation of each of the sector-specific forecasts for the energy-intensive sectors, manufacturing, retail, and hospitality (NAICS 31, 32, 33, 44, & 72). The estimated electricity-related job losses were subtracted from a reference forecast for each sector that simply extrapolated the 20-year average annual growth rate (AGR). The 95% confidence intervals, both with and without robust standard errors, are displayed in gray surrounding the single-point estimations. The delta between the estimate and reference case is the isolated effect of electricity price increases on employment. energy.ky.gov 8

15 Impact on Manufacturing Employment Figure 10: Kentucky Manufacturing Employment Forecast, Of the sectors analyzed, manufacturing, Kentucky's largest economic sector, was the most-responsive sector to changes in electricity prices. Specifically, an increase of 10% in real electricity prices was associated with a reduction of 3.37% in absolute manufacturing employment, and with 95% confidence, between -2.77% and -3.97%. This finding was statistically significant below the level. When using robust standard errors, however, the 95% confidence interval widened to between -0.83% and -5.92% and the significance level dropped to Overall economic activity and time were also significant factors in predicting employment in the manufacturing sector; however, educational attainment as well as the total population levels were not. Time had a statistically significant negative coefficient, reflecting the general trend of contraction of manufacturing both in Kentucky and nationally. Given a 25% increase in real electricity prices by 2025, manufacturing establishments in Kentucky would be expected to permanently shed an additional 17,660 full-time jobs long-run as a direct result of price increases, and with 95% confidence using robust standard errors between 5,764 and 31,022 full-time jobs, ceteris paribus. The manufacturing employment forecast, illustrated in Figure 10 above, was developed by applying the elasticities for the manufacturing sector to the electricity price forecast to estimate electricity price-related job losses, which were subtracted from a baseline forecast developed using the 20-year AGR of -1.16%, and then subtracting predicted historical electricity-related losses, for a net reference AGR of -1.07%. energy.ky.gov 9

16 Impact on Retail Trade Employment Figure 11: Kentucky Retail Trade Employment Forecast, Retail trade, Kentucky's largest employment sector in terms of total employment, was less than half as responsive as the manufacturing sector to increasing electricity prices. Specifically, an increase of 10% in real electricity prices was associated with a reduction of 1.57% in total employment, and with 95% confidence between -1.30% and -1.84%. When using robust standard errors, however, the 95% confidence interval widened between -0.77% and -2.39%. These findings were statistically significant below the level. Education was not a significant factor in determining retail employment; whereas economic activity and total population levels were. Given a 25% increase in real electricity prices by 2025, retail establishments in Kentucky would be expected to fail to create 7,225 full-time jobs long-run, and with 95% confidence using robust standard errors, between 3,916 and 12,160 full-time jobs, ceteris paribus. The retail employment forecast, illustrated in Figure 11 above, was developed by applying the elasticities for the retail sector to the electricity price forecast to estimate electricity price-related job losses, which were subtracted from a baseline forecast developed using the 20-year AGR of %, and then subtracting predicted historical electricity-related losses, for a net reference AGR of %. energy.ky.gov 10

17 Impact on Hospitality Employment Figure 12: Kentucky Hospitality Employment Forecast, Employment in hospitality industries such as restaurants and hotels demonstrated a similar, but weaker, responsiveness as retail employment. Specifically, an increase of 10% in real electricity prices was associated with a reduction of 1.42% in total employment, and with 95% confidence between -1.12% and -1.71%. When using robust standard errors, however, the 95% confidence interval widened between -0.78% and -2.06%. This finding was statistically significant below the level. Education and total population do not appear to be significant factors in determining hospitality sector employment; whereas economic activity and time were both significant. Given a 25% increase in real electricity prices by 2025, restaurants and hotels in Kentucky would be expected to shed 5,352 full-time jobs long-run, and with 95% confidence using robust standard errors, between 2,940 and 7,765 full-time jobs, ceteris paribus. The retail employment forecast, illustrated in Figure 12 above, was developed by applying the elasticities for the retail sector to the electricity price forecast to estimate electricity price-related job losses, which were subtracted from a baseline forecast developed using the 20-year AGR of %. energy.ky.gov 11

18 Impact on Healthcare Employment Employment in the healthcare industry was much less sensitive to increases in electricity prices, and responsiveness was not statistically significant when using robust standard errors. Specifically, a 10% increase in the price of electricity appears to be associated with a 0.43% reduction in overall healthcare employment. However, with 95% confidence and robust standard errors, these effects are not necessarily distinguishable from zero. Healthcare employment was better predicted by educational attainment of the population, overall economic activity, total population levels, and time. Given that the independent variable of interest, real electricity prices, was not significant when using robust standard errors, no forecast for this sector was developed. Impact on Government Employment In government employment, no relationship between electricity prices and total employment could be identified, whereas educational attainment of the population, overall economic activity, and total population levels appeared to have statistically significant effects. Given that the independent variable of interest, real electricity prices, was not significant in any model, no forecast for this sector was developed. Conclusion This study demonstrated that electricity price increases alone may force businesses to seek ways to reduce costs or close, causing substantial job losses in Kentucky s electricity-intensive manufacturing sector, and slowing overall long-term job creation in other sectors. The timing of this transition could exacerbate high unemployment and slow economic growth in the near-term. The Commonwealth s vulnerability to these dynamics could also be worsened if leadership is unaware of them and inadequately prepared for the transition. Kentucky s neighboring states of Indiana, Ohio, and West Virginia exhibit similar vulnerabilities due to the potential for increasing electricity costs and the relative size of their manufacturing sectors. While total employment in the Commonwealth is expected to continue to rise in other sectors, the Commonwealth should maintain focus on education and workforce development in emerging industries that are less reliant on energy-intensive manufacturing processes as well as consider strategies to mitigate vulnerability to price increases and risk exposure. energy.ky.gov 12

19 Data Analyzed Total employment in Kentucky s top five economic sectors, in terms of number of employees as illustrated in Figure 8 on page 5, served as the dependent variables of interest in this study. Total employment by industry was collected from the Bureau of Economic Analysis (BEA) for all 51 entities and all years from 1990 to Data was collected for each state as well as the District of Columbia, in each year, and for each industry, organized by North American Industry Classification System (NAICS) codes. The primary explanatory variable of interest in this study was the natural logarithm of total real electricity price in each state and year expressed in 2010 US$ per kwh. Electricity prices are defined here as the quotient of the total revenue received by electric utilities in state i and in year t divided by the total kilowatt-hours of electricity sold in that state and year. Electricity prices differ from electricity rates, which are only a subset of the total cost and often do not include taxes, environmental surcharges, and fuel costs that vary substantially across time and geography. Thus, electricity prices more accurately reflect the cost for one kilowatt-hour of electricity paid by consumers in a given state and year. This variable was assembled using a variety of datasets from the Energy Information Administration (EIA), including data from the State Energy Data System (SEDS) for years 1990 to 2009 for all states, 10 and where certified data was not yet available using Form EIA and Form EIA-826 for the year The correlation between historical electricity prices derived from Form EIA-861 and EIA-826 to the corresponding certified variables was 0.999; thus, there is almost no difference between the historical data and the 2010 update other than it has not yet been certified and included in SEDS. The following control variables were used: educational attainment, defined as the percentage of the adult population (age 25 years and older) with a bachelor s degree (or higher), collected from the United States Census American Community Survey; population, also collected from the United States Census; state Gross Domestic Product (GDP), collected from the BEA; and year. The following control variables were also tested but ultimately excluded because their effects were not statistically significant: labor force unionization, Standard & Poor s 500 Index, and per capita personal income. There were a total of 51 states included (N=51), the 50 United States as well as the District of Columbia. However, the model s performance would have been improved by ~5% if the District of Columbia had been excluded. All currency variables, namely the price of electricity and State Gross Domestic Product, were adjusted for inflation to 2010 US$ using the Bureau of Labor Statistics (BLS) Consumer Price Index (CPI), which is intended to account for the generally rising cost of goods during this time period. energy.ky.gov 13

20 Analytical Method Using a statistical analysis technique called multiple regression of panel data with fixed effects, this study modeled the responsiveness of employment across the United States to changes in the real price of electricity from 1990 to 2010 for the top five employment sectors in Kentucky: manufacturing (NAICS 31, 32, & 33), retail services (NAICS 44), hospitality (NAICS 72), healthcare (NAICS 62), and government (NAICS 92). Elasticities were developed for each sector to calculate changes in employment given a specific change in the electricity prices and can be generally applied to any state and year. To develop these elasticity coefficients, data were organized into a multidimensional panel, i.e. both time series and cross sectional, enabling simultaneous modeling of the relationships of multiple statistics across both space and time (N x t). Since each observation is non-random, and not independent, for example electricity prices in state i and year t are not independent of prices in state i in year t-1, a fixed effects model was used, which builds upon Ordinary Least Squares (OLS) regression by isolating the time-independent constant difference between states that is correlated with the explanatory variables. Two multiple regression of panel data models with fixed effects, both with and without robust standard errors, were constructed for each of the top five employment sectors in Kentucky, for a total of 10 separate multiple regression models. The multiple regression of panel data model with fixed effects can be generally given by, Where i and t index states and years, such that y it is the dependent variable of interest, employment by industry, in state i in year t, β 0 is the constant y intercept across all states, X is a k by 1 vector of explanatory variables, β j X jit is the product of the observation for each independent variable j through k for state i in year t and the coefficient of X, k is the total number of included independent variables, α i is the time-invariant fixed effect for state i, and ε it are the residuals, and where ε it ~ N(0, σ 2 ), or are approximately normally distributed with a mean of zero. Multiple regression of panel data using fixed effects facilitated controlling for the numerous factors inherently affecting sector-specific employment as well as electricity prices from state to state that have not been accounted for in the independent variables included in this study to isolate the primary national effect of the variable of interest, real electricity prices, on each of the dependent variables, employment by industry. Since this study aims to isolate the unique effect of electricity prices on employment, the model was rerun five times to derive the coefficient for each of the industries of interest by NAICS code. A fixed effects model specifically assumes the existence of unobserved time-invariant heterogeneity, often referred to as unobserved variable bias, which in addition to the included independent variables, is affecting the dependent variable. The fixed effects model will attempt to control for these missing or unobserved between unit (interstate) factors, the fixed effects, to isolate the specific net effect of the independent variables of interest on all units (nationally). The fixed effects model also assumes that these between-unit effects are both time invariant and correlated with the independent variables. A fixed effect model is also functionally, although not computationally, equivalent to assigning an independent indicator energy.ky.gov 14